CN107392131A - A kind of action identification method based on skeleton nodal distance - Google Patents
A kind of action identification method based on skeleton nodal distance Download PDFInfo
- Publication number
- CN107392131A CN107392131A CN201710573316.3A CN201710573316A CN107392131A CN 107392131 A CN107392131 A CN 107392131A CN 201710573316 A CN201710573316 A CN 201710573316A CN 107392131 A CN107392131 A CN 107392131A
- Authority
- CN
- China
- Prior art keywords
- distance
- frame
- bone
- sequence
- skeleton
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Pending
Links
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V40/00—Recognition of biometric, human-related or animal-related patterns in image or video data
- G06V40/20—Movements or behaviour, e.g. gesture recognition
- G06V40/23—Recognition of whole body movements, e.g. for sport training
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/25—Fusion techniques
- G06F18/254—Fusion techniques of classification results, e.g. of results related to same input data
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V20/00—Scenes; Scene-specific elements
- G06V20/40—Scenes; Scene-specific elements in video content
- G06V20/46—Extracting features or characteristics from the video content, e.g. video fingerprints, representative shots or key frames
Landscapes
- Engineering & Computer Science (AREA)
- Theoretical Computer Science (AREA)
- Physics & Mathematics (AREA)
- General Physics & Mathematics (AREA)
- Data Mining & Analysis (AREA)
- Multimedia (AREA)
- Computer Vision & Pattern Recognition (AREA)
- General Engineering & Computer Science (AREA)
- Evolutionary Computation (AREA)
- Evolutionary Biology (AREA)
- Bioinformatics & Computational Biology (AREA)
- Bioinformatics & Cheminformatics (AREA)
- Artificial Intelligence (AREA)
- Health & Medical Sciences (AREA)
- General Health & Medical Sciences (AREA)
- Psychiatry (AREA)
- Social Psychology (AREA)
- Human Computer Interaction (AREA)
- Life Sciences & Earth Sciences (AREA)
- Image Analysis (AREA)
Abstract
The present invention relates to a kind of action identification method based on skeleton nodal distance, step are as follows：(1) three planes of each frame of bone sequence in Descartes's rhombic system are projected respectively, generate skeleton distribution map；(2) the skeletal joint point on skeleton distribution map is extracted；(3) the Euclidean distance between the skeletal joint of each frame of bone sequence is calculated, the distance of all nodes is combined into frame distance vector；(4) by all frame distance vectors of bone sequence composition sequence distance matrix sequentially in time；(5) it is color coded using pseudo- colour coding method two-dimensional matrix of adjusting the distance, obtains color texture figure；(6) picture classification is carried out using deep learning method, so as to complete the motion detection identification mission of the mankind.Action identification method of the invention based on bone nodal distance is not influenceed by environment and interactive action, suitable for interactive action recognition under complex scene, thus has more wide application value.
Description
Technical field
The invention belongs to multimedia signal processing field, is related to computer intelligence, pattern-recognition, machine learning, is by it
A kind of action identification method based on skeleton nodal distance.
Background technology
With the continuous development of computer intellectual technology, in life from now on, human action identification should with wide
Use prospect.Such as：Intelligent monitoring, the somatic sensation television game of man-machine interaction, video frequency searching etc..So research action recognition methods has
Far-reaching researching value.In recent years, with increasingly ripe, easy use, the depth sensing of low cost of computer vision technique
Device, such as Kinect cameras, it is widely used in a variety of applications.Because depth camera is compared with conventional color video camera, there is control
The advantages of bright change is insensitive and can reliably estimate body contour and skeleton, the human action based on depth bone video sequence
Detection identification causes increasing concern.But the stroking technique identification for being currently based on bone is still immature.
Two patents CN106203503A, CN106228109A of application are and the application most phase before present inventor
The publication of pass, its shortcoming are can only to identify the action in single subject scenes, it is impossible to are met interactive dynamic under complex scene
Make the needs identified.
The content of the invention
This method utilizes the skeleton sequence that is obtained by Kinect video camera, by extracting between skeleton node
Range information, it is color coded, and classification based training is carried out to action using deep learning method, it is proposed that one kind is based on human body
The action identification method of artis distance, meet the needs of interactive action recognition under complex scene.
The technical scheme for realizing the object of the invention is：
A kind of action identification method based on skeleton nodal distance, step are as follows：
(1) three planes of each frame of bone sequence in Descartes's rhombic system are projected respectively, generation skeleton distribution
Figure；
(2) the skeletal joint point on skeleton distribution map is extracted；
(3) the Euclidean distance between the skeletal joint of each frame of bone sequence is calculated, the distance of all nodes is combined into framing
Distance vector；
(4) by all frame distance vectors of bone sequence composition sequence distance matrix sequentially in time；
(5) it is color coded using pseudo- colour coding method two-dimensional matrix of adjusting the distance, obtains color texture figure；
(6) picture classification is carried out using deep learning method, so as to complete the motion detection identification mission of the mankind.
Described deep learning method and step is：Using four-way to front, side, top surface, three-dimensional distance colour
Texture maps carry out the parameter training of convolutional neural networks respectively；(2) convolutional neural networks parameter is adjusted；(3) four passages are merged
Action recognition result, the test data of input is classified.To four passage output category probability vector v1, v2, v3, v4Press
Merged according to equation below,
Lable=Fin (max (v1ov2ov3ov4))
Wherein, the element multiplication of o representative vectors, Fin () function stand is using the maximum action of probable value as action recognition
Result.
Concretely comprise the following steps：
The first step, the bone sequence frame number of everything is fixed as t using bilinear interpolation；
Second step, it is assumed that from each frame skeletal graph VxyzM skeleton node of upper extraction, is used
To represent the three dimensional local information of j-th of bone node of the i-th frame；Entirely
In video sequence all bones can be projected to according to three-dimensional information three fronts of three Descartes's rhombic system, side,
Top surface, obtain positive Sxy, side Syz, top surface Sxz, after projection, each frame bone sequence obtains the description of the three black and white action
The picture of bone distribution；
3rd step, in positive Sxy, side Syz, top surface SxzWith skeletal graph VxyzIt is upper to obtain bone node respectivelyArrive other
The Euclidean distance of m-1 bone node
djk i=| | Pj i-Pk i||2,j∈1,...,m-1；k∈j+1,...,m
4th step, each a total of m of frame (m-1)/2 bone node pair, therefore m (m-1)/2 distance is obtained, again to it
Marked asWherein, l ∈ 1 ..., m (m-1)/2, by this m (m-1)/2 bone nodal distance component frame distance vector Di：
Di={ d1 i,d2 i,...,dl i}
5th step, the frame Distance matrix D of each frameiCombine sequentially in time, obtain sequence distance matrix H：
H={ D1,D2,...,Dt}
6th step, pseudo- color coding is carried out to sequence distance matrix, obtains positive Sxy, side Syz, top surface Sxz, it is three-dimensional away from
From VxyzColor texture figure.
The advantages of the present invention：
1st, the action identification method of the invention based on bone nodal distance is not influenceed by environment and interactive action, suitable for multiple
Interactive action recognition under miscellaneous scene, thus there is more wide application value.
2nd, the present invention saves the temporal characteristics acted in bone sequence by recording the change of the distance between bone node.
The difference of distance can correspond to different actions between bone node so that motion characteristic is more obvious, more easily nerve net
Network learns, so as to improve the accuracy rate of action recognition.
3rd, the present invention uses four passages, and the feature extracted is trained using deep learning technology, finally used
Integration technology, the classification results of four passages are merged, realize the identification function of objective and accurate and efficient human action.
4th, the crucial bone node of the invention by extracting fixed position on each frame of bone sequence, so as to effective
Extract space characteristics, compared with traditional-handwork extracts motion characteristic, there is objectivity, general applicability, reduce randomness with
The interference of subjectivity selection.So that recognition result is more objective credible, application is stronger.
Brief description of the drawings
Fig. 1 is the entirely action recognition frame diagram based on skeleton node.
Embodiment
Below in conjunction with the accompanying drawings and the invention will be further described by specific embodiment, and following examples are descriptive
, it is not limited, it is impossible to which protection scope of the present invention is limited with this.
A kind of action identification method based on skeleton nodal distance, step are as follows：
1) mapping of the bone sequence to picture
Assuming that there is a series of bone sequence of human actions, now action is identified, the bone generally each acted
The frame number t of sequencexIt is uncertain.
The first step, the bone sequence frame number of everything is fixed as t using bilinear interpolation.
Second step, it is assumed that from each frame skeletal graph VxyzM skeleton node of upper extraction, is used
To represent the three dimensional local information of j-th of bone node of the i-th frame.Entirely
All bones can project in three planes of three Descartes's rhombic system (just according to three-dimensional information in video sequence
Face, side, top surface), obtain positive Sxy, side Syz, top surface Sxz.After such projection, each frame bone sequence can obtain three
Open the picture for describing action bone distribution of black and white.With abundant spatial information, but retouching for temporal information is not had
State.
3rd step, in positive Sxy, side Syz, top surface SxzWith skeletal graph VxyzIt is upper to obtain bone node respectivelyArrive other
The Euclidean distance of m-1 bone node
djk i=| | Pj i-Pk i||2,j∈1,...,m-1；k∈j+1,...,m
4th step, each a total of m of frame (m-1)/2 bone node pair, therefore m (m-1)/2 distance is obtained, again to it
Marked asWherein, l ∈ 1 ..., m (m-1)/2, by this m (m-1)/2 bone nodal distance component frame distance vector Di：
Di={ d1 i,d2 i,...,dl i}
5th step, the frame Distance matrix D of each frameiCombine sequentially in time, obtain sequence distance matrix H：
H={ D1,D2,...,Dt}
6th step, pseudo- color coding is carried out to sequence distance matrix, obtains color texture figure.Four groups of colours can now be obtained
Texture maps, i.e., positive Sxy, side Syz, top surface Sxz, three-dimensional distance VxyzColor texture figure.
2) picture classification is carried out using deep learning, so as to complete the motion detection identification mission of the mankind.
Herein, we under linux system, are led to using current public deep learning framework caffe
NvidiaGTXTITANX video cards carry out the acceleration of convolutional neural networks training, using the network structure of AlexNet in the tool box
Color texture picture Jing Guo color coding is trained.
The first step, using the training dataset in data set, with four passages to positive Sxy, side Syz, top surface Sxz, three
Tie up distance VxyzColor texture figure carry out the parameter trainings of convolutional neural networks respectively.
Second step, using fine setting model optimization network parameter.
3rd step, to four passage output category probability vector v1, v2, v3, v4Merged according to equation below,
Lable=Fin (max (v1ov2ov3ov4))
Wherein, the element multiplication of o representative vectors, Fin () function stand is using the maximum action of probable value as action recognition
Result.
Experimental result explanation for the present invention on the data sets such as NTU RGB-D, UTD-MHAD below：
The Realization of Simulation of algorithm is carried out on Matlab-2014b platforms, has calculated training dataset and test data
Collect in front, side, top surface and bone nodal distance figure spatially, and color texture is converted into using pseudo- color coding
Figure.We under linux system, pass through NvidiaGTXTITANX video cards using current public deep learning framework caffe
Carry out the acceleration of convolutional neural networks training.Still completed on Matlab-2014b platforms last probability vector fusion part.
This method is concentrated in internationally recognized human action's identification data and tested, and is instructed in data set NTU RGB-D
The method of salary distribution for practicing collection and test set uses cross subject and cross the view methods of salary distribution.Test result is as follows：
Acted comprising 60 classes on the NTU RGB-D data sets (wherein comprising 11 interactive actions), obtain 76.2% He respectively
82.3% recognition accuracy；The method of salary distribution of training set and test set uses cross in data set UTD-MHAD
The subject methods of salary distribution.Test result is：On the UTD-MHAD data sets acted comprising 27 classes, 88.10% knowledge is obtained
Other accuracy rate.The experimental result contrasts detects recognizer with the human action leading in the world based on bone sequence, obtains
It is higher than the classification degree of accuracy of remaining algorithm.
Above-described is only the preferred embodiment of the present invention, it is noted that for one of ordinary skill in the art
For, on the premise of inventive concept is not departed from, various modifications and improvements can be made, these belong to the protection of the present invention
Scope.
Claims (4)
1. a kind of action identification method based on skeleton nodal distance, step are as follows：
(1) three planes of each frame of bone sequence in Descartes's rhombic system are projected respectively, generate skeleton distribution map；
(2) the skeletal joint point on skeleton distribution map is extracted；
It is characterized in that：
Calculate each frame of bone sequence skeletal joint between Euclidean distance, by the distance of all nodes be combined into frame pitch from
Vector；
(4) by all frame distance vectors of bone sequence composition sequence distance matrix sequentially in time；
(5) it is color coded using pseudo- colour coding method two-dimensional matrix of adjusting the distance, obtains color texture figure；
(6) picture classification is carried out using deep learning method, so as to complete the motion detection identification mission of the mankind.
2. the action identification method according to claim 1 based on skeleton nodal distance, it is characterised in that：Described
Deep learning method and step is：(1) the color texture figure of front, side, top surface, three-dimensional distance is carried out respectively using four-way
The parameter training of convolutional neural networks；(2) convolutional neural networks parameter is adjusted；(3) the action recognition result of four passages is merged, it is right
The test data of input is classified.
3. the action identification method according to claim 1 based on skeleton nodal distance, it is characterised in that：Specific step
Suddenly it is：
The first step, the bone sequence frame number of everything is fixed as t using bilinear interpolation；
Second step, it is assumed that from each frame skeletal graph VxyzM skeleton node of upper extraction, is used
To represent the three dimensional local information of j-th of bone node of the i-th frame；Whole video
All bones can project to three fronts, side, top surfaces of three Descartes's rhombic system according to three-dimensional information in sequence,
Obtain positive Sxy, side Syz, top surface Sxz, after projection, each frame bone sequence obtains the description of three black and white action bone point
The picture of cloth；
3rd step, in positive Sxy, side Syz, top surface SxzWith skeletal graph VxyzIt is upper to obtain bone node P respectivelyj iTo other m-1
The Euclidean distance d of bone nodejk i：
djk i=| | Pj i-Pk i||2,j∈1,...,m-1；k∈j+1,...,m
4th step, each a total of m of frame (m-1)/2 bone node pair, therefore m (m-1)/2 distance is obtained, again to its label
For dli：Wherein, l ∈ 1 ..., m (m-1)/2, by this m (m-1)/2 bone nodal distance component frame distance vector Di：
Di={ d1 i,d2 i,...,dl i}
5th step, the frame Distance matrix D of each frameiCombine sequentially in time, obtain sequence distance matrix H：
H={ D1,D2,...,Dt}
6th step, pseudo- color coding is carried out to sequence distance matrix, obtains positive Sxy, side Syz, top surface Sxz, three-dimensional distance Vxyz
Color texture figure.
4. the action identification method according to claim 2 based on skeleton nodal distance, it is characterised in that：To four
Passage output category probability vector v1, v2, v3, v4Merged according to equation below,
Wherein,The element multiplication of representative vector, knot of the Fin () function stand using the maximum action of probable value as action recognition
Fruit.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201710573316.3A CN107392131A (en) | 2017-07-14 | 2017-07-14 | A kind of action identification method based on skeleton nodal distance |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201710573316.3A CN107392131A (en) | 2017-07-14 | 2017-07-14 | A kind of action identification method based on skeleton nodal distance |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN107392131A true CN107392131A (en) | 2017-11-24 |
Family
ID=60340130
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201710573316.3A Pending CN107392131A (en) | 2017-07-14 | 2017-07-14 | A kind of action identification method based on skeleton nodal distance |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN107392131A (en) |
Cited By (11)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN108257139A (en) * | 2018-02-26 | 2018-07-06 | 中国科学院大学 | RGB-D three-dimension object detection methods based on deep learning |
| CN108898064A (en) * | 2018-06-05 | 2018-11-27 | 华东师范大学 | A kind of 3D bone action identification method based on spatial match |
| CN109614899A (en) * | 2018-11-29 | 2019-04-12 | 重庆邮电大学 | A kind of human motion recognition method based on Lie group feature and convolutional neural networks |
| CN109670401A (en) * | 2018-11-15 | 2019-04-23 | 天津大学 | A kind of action identification method based on skeleton motion figure |
| CN109685037A (en) * | 2019-01-08 | 2019-04-26 | 北京汉王智远科技有限公司 | A kind of real-time action recognition methods, device and electronic equipment |
| CN110490034A (en) * | 2018-05-14 | 2019-11-22 | 欧姆龙株式会社 | Motion analysis device, action-analysing method, recording medium and motion analysis system |
| CN110717384A (en) * | 2019-08-30 | 2020-01-21 | 深圳大学 | Video interactive behavior recognition method and device |
| CN111223549A (en) * | 2019-12-30 | 2020-06-02 | 华东师范大学 | Mobile end system and method for disease prevention based on posture correction |
| WO2020107847A1 (en) * | 2018-11-28 | 2020-06-04 | 平安科技（深圳）有限公司 | Bone point-based fall detection method and fall detection device therefor |
| CN115131301A (en) * | 2022-06-15 | 2022-09-30 | 北京长木谷医疗科技有限公司 | Method and system for intelligently identifying osteoarthritis based on deep learning |
| CN115457518A (en) * | 2022-08-30 | 2022-12-09 | 淮阴工学院 | Driver behavior recognition method and system based on attitude perception and geometric constraint |
Citations (12)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN103529944A (en) * | 2013-10-17 | 2014-01-22 | 合肥金诺数码科技股份有限公司 | Human body movement identification method based on Kinect |
| CN105138995A (en) * | 2015-09-01 | 2015-12-09 | 重庆理工大学 | Time-invariant and view-invariant human action identification method based on skeleton information |
| US20160086350A1 (en) * | 2014-09-22 | 2016-03-24 | Foundation for Research and Technology - Hellas (FORTH) (acting through its Institute of Computer | Apparatuses, methods and systems for recovering a 3-dimensional skeletal model of the human body |
| CN105930767A (en) * | 2016-04-06 | 2016-09-07 | 南京华捷艾米软件科技有限公司 | Human body skeleton-based action recognition method |
| CN106156714A (en) * | 2015-04-24 | 2016-11-23 | 北京雷动云合智能技术有限公司 | The Human bodys' response method merged based on skeletal joint feature and surface character |
| CN106203363A (en) * | 2016-07-15 | 2016-12-07 | 中国科学院自动化研究所 | Human skeleton motion sequence Activity recognition method |
| CN106203503A (en) * | 2016-07-08 | 2016-12-07 | 天津大学 | A kind of action identification method based on skeleton sequence |
| CN106845386A (en) * | 2017-01-16 | 2017-06-13 | 中山大学 | A kind of action identification method based on dynamic time warping Yu Multiple Kernel Learning |
| US20170168586A1 (en) * | 2015-12-15 | 2017-06-15 | Purdue Research Foundation | Method and System for Hand Pose Detection |
| CN106897714A (en) * | 2017-03-23 | 2017-06-27 | 北京大学深圳研究生院 | A kind of video actions detection method based on convolutional neural networks |
| CN106897670A (en) * | 2017-01-19 | 2017-06-27 | 南京邮电大学 | A kind of express delivery violence sorting recognition methods based on computer vision |
| CN106909887A (en) * | 2017-01-19 | 2017-06-30 | 南京邮电大学盐城大数据研究院有限公司 | A kind of action identification method based on CNN and SVM |
-
2017
- 2017-07-14 CN CN201710573316.3A patent/CN107392131A/en active Pending
Patent Citations (12)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN103529944A (en) * | 2013-10-17 | 2014-01-22 | 合肥金诺数码科技股份有限公司 | Human body movement identification method based on Kinect |
| US20160086350A1 (en) * | 2014-09-22 | 2016-03-24 | Foundation for Research and Technology - Hellas (FORTH) (acting through its Institute of Computer | Apparatuses, methods and systems for recovering a 3-dimensional skeletal model of the human body |
| CN106156714A (en) * | 2015-04-24 | 2016-11-23 | 北京雷动云合智能技术有限公司 | The Human bodys' response method merged based on skeletal joint feature and surface character |
| CN105138995A (en) * | 2015-09-01 | 2015-12-09 | 重庆理工大学 | Time-invariant and view-invariant human action identification method based on skeleton information |
| US20170168586A1 (en) * | 2015-12-15 | 2017-06-15 | Purdue Research Foundation | Method and System for Hand Pose Detection |
| CN105930767A (en) * | 2016-04-06 | 2016-09-07 | 南京华捷艾米软件科技有限公司 | Human body skeleton-based action recognition method |
| CN106203503A (en) * | 2016-07-08 | 2016-12-07 | 天津大学 | A kind of action identification method based on skeleton sequence |
| CN106203363A (en) * | 2016-07-15 | 2016-12-07 | 中国科学院自动化研究所 | Human skeleton motion sequence Activity recognition method |
| CN106845386A (en) * | 2017-01-16 | 2017-06-13 | 中山大学 | A kind of action identification method based on dynamic time warping Yu Multiple Kernel Learning |
| CN106897670A (en) * | 2017-01-19 | 2017-06-27 | 南京邮电大学 | A kind of express delivery violence sorting recognition methods based on computer vision |
| CN106909887A (en) * | 2017-01-19 | 2017-06-30 | 南京邮电大学盐城大数据研究院有限公司 | A kind of action identification method based on CNN and SVM |
| CN106897714A (en) * | 2017-03-23 | 2017-06-27 | 北京大学深圳研究生院 | A kind of video actions detection method based on convolutional neural networks |
Non-Patent Citations (2)
| Title |
|---|
| CHUANKUN LI.ET: "Joint Distance Maps Based Action Recognition With Convolutional Neural Networks", 《IEEE SIGNAL PROCESSING LETTERS》 * |
| 陈渊博: "视频序列中的人体动作识别", 《中国博士学位论文全文数据库 信息科技辑》 * |
Cited By (16)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN108257139A (en) * | 2018-02-26 | 2018-07-06 | 中国科学院大学 | RGB-D three-dimension object detection methods based on deep learning |
| CN108257139B (en) * | 2018-02-26 | 2020-09-08 | 中国科学院大学 | RGB-D three-dimensional object detection method based on deep learning |
| CN110490034A (en) * | 2018-05-14 | 2019-11-22 | 欧姆龙株式会社 | Motion analysis device, action-analysing method, recording medium and motion analysis system |
| CN108898064A (en) * | 2018-06-05 | 2018-11-27 | 华东师范大学 | A kind of 3D bone action identification method based on spatial match |
| CN109670401A (en) * | 2018-11-15 | 2019-04-23 | 天津大学 | A kind of action identification method based on skeleton motion figure |
| WO2020107847A1 (en) * | 2018-11-28 | 2020-06-04 | 平安科技（深圳）有限公司 | Bone point-based fall detection method and fall detection device therefor |
| CN109614899A (en) * | 2018-11-29 | 2019-04-12 | 重庆邮电大学 | A kind of human motion recognition method based on Lie group feature and convolutional neural networks |
| CN109614899B (en) * | 2018-11-29 | 2022-07-01 | 重庆邮电大学 | Human body action recognition method based on lie group features and convolutional neural network |
| CN109685037A (en) * | 2019-01-08 | 2019-04-26 | 北京汉王智远科技有限公司 | A kind of real-time action recognition methods, device and electronic equipment |
| CN109685037B (en) * | 2019-01-08 | 2021-03-05 | 北京汉王智远科技有限公司 | Real-time action recognition method and device and electronic equipment |
| CN110717384A (en) * | 2019-08-30 | 2020-01-21 | 深圳大学 | Video interactive behavior recognition method and device |
| CN111223549A (en) * | 2019-12-30 | 2020-06-02 | 华东师范大学 | Mobile end system and method for disease prevention based on posture correction |
| CN115131301A (en) * | 2022-06-15 | 2022-09-30 | 北京长木谷医疗科技有限公司 | Method and system for intelligently identifying osteoarthritis based on deep learning |
| WO2023241032A1 (en) * | 2022-06-15 | 2023-12-21 | 北京长木谷医疗科技有限公司 | Deep learning-based method and system for intelligently identifying osteoarthritis |
| CN115457518A (en) * | 2022-08-30 | 2022-12-09 | 淮阴工学院 | Driver behavior recognition method and system based on attitude perception and geometric constraint |
| CN115457518B (en) * | 2022-08-30 | 2024-01-26 | 淮阴工学院 | Driver behavior recognition method and system based on gesture perception and geometric constraint |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN107392131A (en) | A kind of action identification method based on skeleton nodal distance | |
| CN108932500B (en) | A kind of dynamic gesture identification method and system based on deep neural network | |
| CN107358257B (en) | Under a kind of big data scene can incremental learning image classification training method | |
| CN106778604B (en) | Pedestrian re-identification method based on matching convolutional neural network | |
| CN108830252A (en) | A kind of convolutional neural networks human motion recognition method of amalgamation of global space-time characteristic | |
| CN106203503B (en) | A kind of action identification method based on bone sequence | |
| CN109961006A (en) | A kind of low pixel multiple target Face datection and crucial independent positioning method and alignment schemes | |
| CN105160310A (en) | 3D (three-dimensional) convolutional neural network based human body behavior recognition method | |
| CN107292813A (en) | A kind of multi-pose Face generation method based on generation confrontation network | |
| CN112800906B (en) | Improved YOLOv 3-based cross-domain target detection method for automatic driving automobile | |
| CN107808129A (en) | A kind of facial multi-characteristic points localization method based on single convolutional neural networks | |
| CN109543632A (en) | A kind of deep layer network pedestrian detection method based on the guidance of shallow-layer Fusion Features | |
| CN112686928B (en) | Moving target visual tracking method based on multi-source information fusion | |
| CN108898063A (en) | A kind of human body attitude identification device and method based on full convolutional neural networks | |
| Li et al. | Sign language recognition based on computer vision | |
| CN110728183A (en) | Human body action recognition method based on attention mechanism neural network | |
| CN112347861A (en) | Human body posture estimation method based on motion characteristic constraint | |
| CN106228109A (en) | A kind of action identification method based on skeleton motion track | |
| CN109858407A (en) | A kind of video behavior recognition methods based on much information stream feature and asynchronous fusion | |
| CN105069745A (en) | face-changing system based on common image sensor and enhanced augmented reality technology and method | |
| CN110163567A (en) | Classroom roll calling system based on multitask concatenated convolutional neural network | |
| CN111881716A (en) | Pedestrian re-identification method based on multi-view-angle generation countermeasure network | |
| CN109670401B (en) | Action recognition method based on skeletal motion diagram | |
| CN113128424A (en) | Attention mechanism-based graph convolution neural network action identification method | |
| Estrela et al. | Sign language recognition using partial least squares and RGB-D information |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| PB01 | Publication | ||
| PB01 | Publication | ||
| SE01 | Entry into force of request for substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| RJ01 | Rejection of invention patent application after publication | ||
| RJ01 | Rejection of invention patent application after publication |
Application publication date: 20171124 |